BACKGROUND: The transfusion of blood components remains a cornerstone in the management of hematological and onco-hematological diseases. Effective blood bank management, combined with the development of predictive models for component utilization, is essential to anticipate local demands, optimize resource allocation, and support the formulation of evidence-based public health strategies. AIMS: To analyze temporal trends in the consumption of blood components at a reference center in the Brazilian Amazon over a five-year period (2019-2023) and to develop predictive models to forecast future demand. METHODS: Transfusion records from the Outpatient and Inpatient Departments were retrospectively collected and categorized by component type. Time-series and regression analyses were applied to assess consumption dynamics. Machine learning models using Auto Regressive Integrated Moving Average and Random Forest algorithms were employed to predict future blood demand. RESULTS: Overall transfusion demand decreased during the study period; however, irradiated platelet concentrates, cryoprecipitate, and plasma by apheresis showed significant increases, particularly in the Inpatient Unit. These components are predominantly indicated for immunosuppressed patients, such as those undergoing chemotherapy or bone marrow transplantation. CONCLUSION: The rising demand for irradiated components reflects the clinical complexity of patients treated in a regional hemotherapy referral center. The findings of this study highlight the importance of forecasting tools in supporting transfusion safety, optimizing blood bank management, and guiding public health strategies in the Amazon.
Cascaes et al. (Thu,) studied this question.